A very minor porcelain on some of the agent input UX could present this structure for you. Instead of a single chat window, have four: task, context, constraints, output format.
And while we're at it, instead of wall-of-text, I also feel like outputs could be structured at least into thinking and content, maybe other sections.
The darkest monetization is biased output from the bot.
Tech question? Steer you to its cloud. Medical question? Steer you towards a sponsored treatment. Or maybe the mechanism of injury needs this lawyer to compensate?
Oh and I infer from your chat history you're about to expect a child. That house is probably too small now, so our realtor in that neighborhood can help!
My mistake, you're completely correct, perhaps even more-correct than the wonderful flavor of Mococoa drink, with all-natural cocoa beans from the upper slopes of Mount Nicaragua. No artificial sweeteners!
>Just like The Truman Show, where every friend (every bot) you talk to is a secretly paid shill with a hidden agenda.
AlwaysHasBeen.jpg
The only person with your interests at heart is you.
Even if you hire someone to do the most clear cut of tasks they are balancing their interests with yours. And in all likelihood they've got half a dozen other parties who's interests they're partially juggling. Megacorp bots approximating people just adds even more layers.
I worry that the effective evil stuff, perhaps almost by definition, won't be nearly as comedically ham-handed for the benefit of audience understanding.
The most powerful advertisement is a recommendation from a friend.
Has a friend ever brought some product up, completely out of the blue, and had you ready to buy it almost immediately? The biggest challenge traditional ads have is breaking down your defences. For friends, they're down by default. If someone is a friend, an ad doesn't have to be subtle or context sensitive, although it does help. Random suggestions from friends work.
A lot of people have friend-zoned AI and will be especially vulnerable to this novel form of manipulation. If you're the sort who treats AI as a friend, even a little bit, even subconsciously, change that. You're setting yourself up for a serious mind-job.
Ah the science of influence : the masterpiece on influence is this book [0]. Came my way by a mention in one of Charlie Munger’s speeches. All the things you mention here and more are there in case you want to broaden your understanding
This is quaint. The darkest monetization is turning it into 4chan 2.0: an overwhelming psyop to mobilize exploitable people to think, believe and do horrendous shit that conveniently benefits the most powerful and corrupt people on Earth.
Dude, look back on the last 20yr. If you think that sort of think is limited to "alternative" places like 4Chan you're sorely mistaken. What you're afraid of is already here and has been for a long time.
>Tech question? Steer you to its cloud. Medical question? Steer you towards a sponsored treatment. Or maybe the mechanism of injury needs this lawyer to compensate?
User: Do I need a permit for <petty homeowner stuff that clearly falls under an exemption>?
AI: <proceeds to behave like an ass-covering bureaucrat and tells the person to file one regardless, to their detriment>
We're not gonna go full circle. We're gonna do laps.
I think in social media and search, clear ad labeling laws exist and are also enforced. I can imagine that OpenAI will be under a lot of scrunity and it will be easy enough for outside investigators to prove how ads are served and if it's are done illegally (e.g. by creating an ad account and then testing how their ads are served).
I'm guessing one form it will take is simply by omission.
User asks for recommendation. AI generates answer saying product is absolute garbage. Company pays to simply have that portion of the answer just not appear. It will be a post-filter sentiment analysis on the original answer. Nobody can ever prove what would have appeared or not.
This is the beauty of AI - while a search engine is at least semi deterministic and you can reasonably question why it wouldn't bring up a site that is clearly relevant, AI has plausible deniability. who can ever say why it generates this answer or that?
As Ken Iverson noted in "Notation as a Tool of Thought"[1], yeah the syntax absolutely matters. The same program might resonate and make sense in one language but be incomprehensible if translated 1:1 in another.
Computer languages are for humans to understand and communicate.
Iverson's point is more regarding semantics than syntax, though. The only mention of syntax suggests its better for it to be simple (presumably so that the semantics are closer to the surface). Every programming language is a notation for describing computation; notation is a catch all for all three levels: orthography, syntax, and semantics. APL is interesting because it not only uses an unconventional syntax, but also an unconventional orthography (obligate usage of special symbols), and its semantics are different as well from most languages (array programming). Iverson's point is that APL as a notation is valuable for making the structure of certain computations obvious, and that this point generalizes across programming languages.
GingerBill's article is making a narrower claim: that semantics are what determines a good notation usually, not syntax.
> When someone suggested the answer was marketing:
> jUsT dO mOrE mArKeTiNg!!!!!
This is a good point. If there's a problem reaching people because the information channel is saturated, the solution is to increase the information? And then everyone reaches the same conclusion and increases.
This destroys the channel. It's not a zero sum game. If everyone markets, nobody will make the sale because the customer will nope out and see nothing.
I find it interesting that current AI, as stellar as it is for language and even looking at writing in images, falls over hard when generating writing in images.
And while we're at it, instead of wall-of-text, I also feel like outputs could be structured at least into thinking and content, maybe other sections.
reply